2015 PACE NSC Seeding/General Seeding Discussion

[AKKOLADE]

What I will and won't be doing in this thread:

Will:

Discuss the results of seeding
Solicit advice regarding how to improve seeding
Reviewing the process of how I did seeding (in a later post)
Giving advice on how to address seeding for your tournament

Won't:

Reveal specifics about individual teams' seeding
Get into arguments about individual teams' seeding
Discuss tournament format

[AKKOLADE]

It was a busy day at the hospital, but here's some numbers I managed to come up with on breaks and the like.

Focusing on prelim groups for now, since that's what the seeding had the most impact on:

Average PPG per team by group

Shepard - 325.7125 PPG
Conrad - 318.575 PPG
Cernan - 318.025 PPG
Bean - 316.075 PPG
Armstrong - 315.525 PPG
Scott - 315.175 PPG
Schmitt - 313.925 PPG
Irwin - 311.425 PPG
Mitchell - 306.0625 PPG
Aldrin - 303.75 PPG
Young - 308.025 PPG
Duke - 300.3625 PPG

Seeding accuracy

Teams were seeded 1-96 and divided into groups as evenly as possible. The highest seeded team in their bracket would be on line 1, second highest seeded would be on line 2, etc. If a team was on line 1 and finished first in their group, that would be considered a "perfect line." If they instead finished second, their line for the purposes of this would be off by one.

Of the 96 teams:
51 teams finished "perfectly" on their seed line in the preliminary bracket.
33 teams finished within one seed line
8 teams finished within two seed lines
3 teams finished within three seed lines
1 team finished within four seed lines - this was LASA B with their travel issues.

I'll go into more detail on things when I'm not watching the NBA Finals.

[mushroom]

You'd might expect a prelim bracket with more underseeded teams to be objectively stronger since their strength was underestimated. Since each prelim bracket sends two teams to each tier, you would expect underseeded teams to perform well on Saturday afternoon because they were pushed into a lower tier by teams they shouldn't have been grouped with. To look for incorrectly seeded or particularly strong/weak prelim brackets, I looked at the Saturday afternoon results (combined W/L and average PPG - they are highly correlated) of the teams from each prelim bracket. There was one (Scott) that stood out as being particularly strong, and a few that were pretty weak.

This doesn't do much to identify individual teams that might have been seeded poorly. However, one method I tried to see if brackets were uniformly strong/weak or just had one or two particularly strong/weak teams was, for each prelim bracket, to take each team's average margin of victory in the morning and find the standard deviation of those average margins of victory for all of the teams in that prelim bracket. Excepting Scott, there was a moderate trend (r^2 = .38) that the stronger prelim brackets had closer average margins of victory than the weaker prelim brackets, so the weak brackets may have suffered from the lack of a strong group leader or their 7th/8th seeds while the strong brackets were more uniformly strong.

I haven't yet found a way that would've predicted any of this that could improve the seeding process, but here's some data that might help someone get there. I don't know if any of this is statistically sound, but it was fun regardless.

Prelim Bracket's Saturday Afternoon Record and Average PPG
Scott: 26-14 (338 PPG)
Aldrin: 24-16 (315 PPG)
Conrad: 23-17 (310 PPG)
Cernah: 23-17 (300 PPG)
Shepard: 22-18 (329 PPG)
Mitchell: 21-19 (319 PPG)
Irwin: 21-19 (310 PPG)
Bean: 19-21 (297 PPG)
Duke: 17-23 (280 PPG)
Young: 16-24 (269 PPG)
Schmitt: 15-25 (283 PPG)
Armstrong: 13-27 (270 PPG)

[AKKOLADE]

That's really interesting stuff. I look forward to looking it over more.

[Northern Central Railway]

Seeding accuracy

Teams were seeded 1-96 and divided into groups as evenly as possible. The highest seeded team in their bracket would be on line 1, second highest seeded would be on line 2, etc. If a team was on line 1 and finished first in their group, that would be considered a "perfect line." If they instead finished second, their line for the purposes of this would be off by one.

Of the 96 teams:
51 teams finished "perfectly" on their seed line in the preliminary bracket.
33 teams finished within one seed line
8 teams finished within two seed lines
3 teams finished within three seed lines
1 team finished within four seed lines - this was LASA B with their travel issues.

Regarding this, would it be possible to see the total number of lines away from projected finishes each prelim bracket broke?

[AKKOLADE]

Overall? No. That'd be a bit too exact and, given other reactions so far (from before this thread), I don't think that'd be the best path to head down. I will probably talk specifically about major "misses" and how to handle those.

Planning a bigger general post in the future, but I have a week and a half of a rotation that's left and it's going to eat up all my time for now.